GNU Radio the story so far...

After experimenting with a standalone C program, I decided to implement everything in GNU Radio. The hope is that eventually we can get something running in real-time, and this is the first step towards that goal.

So, at this point the 'energy detector' is implemented in two parts: ed.py and a modified version of gr_bin_statistics_f.

ed.py

This Python script follows much of the GNU Radio examples and uses the USRP as a signal source. Currently the script allows you to collect N samples, then perform an FFT followed by squaring the magnitude output. Originally the time-averaging of the PSDs was done in gr_bin_statistics_f, but the same if-not-better results were obtained using the pre-built single_pole_iir_filter. Finally the output is scaled with a log function before an output file is generated with the test statistics.

Here is the GNU Radio flow-graph:

USRP -> s2v -> FFT -> c2mag -> avg -> log -> stats

gr_bin_statistics_f

Originally, gr_bin_statistics_f was a key component in the usrp_spectrum_sense.py script. It's original function was to gather the maximum values of the FFT output and generate messages for the Python code. It implements a simple state-machine which also allows you to callback to the Python code and re-tune the USRP.

In it's current heavily-modified state, it simple calculates an average of the FFT output over a given window to generate our test stats. At this point I intend on creating a whole new block which removes much of the old functionality.

So, given a window center frequency and bandwidth, the modified code sums M bins and generates a test stat T in the form of a Python message. These messages are then tallied up by ed.py and put into an output file which is plotted directly with GNUplot.

---

In the end we have essentially the same functionality as my stand-alone C program, only now everything is implemented under GNU Radio and we are much closer to something that can be tuned to run in real-time.